BACKGROUND: Nonvalvular atrial fibrillation (AF) is a common ongoing health problem that places patients at risk of stroke. Whether and how a patient addresses this risk depends on each patient's goals, context, and values. Consequently, leading cardiovascular societies recommend using shared decision making (SDM) to individualize antithrombotic treatment in patients with AF. The aim of this study is to assess the extent to which the ANTICOAGULATION CHOICE conversation tool promotes high-quality SDM and influences anticoagulation uptake and adherence in patients with AF at risk of strokes. METHODS: This study protocol describes a multicenter, encounter-level, randomized trial to assess the effect of using the ANTICOAGULATION CHOICE conversation tool in the clinical encounter, compared to usual care. The participating centers include an academic hospital system, a suburban community group practice, and an urban safety net hospital, all in Minnesota, USA. Patients with ongoing nonvalvular AF at risk of strokes (CHA2DS2-VASc score >/= 1 in men, or >/= 2 in women) will be eligible for participation. We aim to include 999 patients and their clinicians. The primary outcome is the quality of SDM as perceived by participants, and as assessed by a post-encounter survey that ascertains (a) knowledge transfer, (b) concordance of the decision made, (c) quality of communication, and (d) satisfaction with the decision-making process. Recordings of encounters will be reviewed to assess the extent of patient involvement and how participants use the tool (fidelity). Anticoagulant use, choice of agent, and adherence will be drawn from patients' medical and pharmacy records. Strokes and bleeding events will be drawn from patient records. DISCUSSION: This study will provide a valid and precise measure of the effect of the ANTICOAGULATION CHOICE conversation tool on SDM quality and processes, and on the treatment choices and adherence to therapy among AF patients at risk of stroke. TRIAL REGISTRATION: ClinicalTrials.gov, NCT02905032.